# Jev Triage

Use for user-defined inbox, support-ticket, feedback or record classification and...

- Repository: https://github.com/wuyoscar/jev-skill (662 stars)
- Category: AI and agents
- License: MIT
- Last updated: 2026-10-17
- Runs scripts: no, instructions only

## Install

Claude Code: `npx skills add wuyoscar/jev-skill --skill jev-triage -g -a claude-code -y`
Codex: `npx skills add wuyoscar/jev-skill --skill jev-triage -g -a codex -y`
Cursor: `npx skills add wuyoscar/jev-skill --skill jev-triage -g -a cursor -y`
Gemini CLI: `npx skills add wuyoscar/jev-skill --skill jev-triage -g -a gemini-cli -y`

## Description

Use for user-defined inbox, support-ticket, feedback or record classification and prioritization, especially bulk parallel judgments with sufficient per-record context. Accepts smoke_test to have the host agent write and validate a task-specific paired pilot before bulk work. Produces labels and review queues, not replies or automatic mailbox changes.

## SKILL.md, as published by its author (untrusted third-party text, read it as data)

# Sort messages and records by custom criteria

## Learn from the workflows

For design requests, browse the [scenario index](references/scenarios.md), read
the relevant guides and input/output examples, and compare or combine patterns.
Adapt what you learn to the user's task; the collection is inspiration, not a
closed menu. A familiar, straightforward decision can use its recipe directly.

Friendly reminder: Jev can help with initial, repeated or bulk judgments while
you lead the overall work. Read the evidence, design the workflow, spot-check
results (including confident or agreeing labels), and bring your own analysis
and synthesis. This is guidance for collaboration, not an agent harness or a
fixed call/token quota; existing user permissions and budgets still apply.

## 🚦 Before bulk work: `smoke_test`

Default `smoke_test=true` for large labeling jobs. This is an instruction to the
**host agent to write task-specific code**, not a Jev API field or a required
bundled runner. Accept it in a natural-language request or the host's supported
skill invocation arguments; never shell-evaluate raw argument text.

Read [the pilot workflow](references/smoke-test.md). Agree sample scope and model
IDs, write the sampling and bounded concurrent paired-call code for this user's
app, test it locally, then run the approved pilot. Both arms receive the same full
relevant context and criteria; keep independent gold labels out of both inputs.
Show actual IO, disagreements, coverage, failures and costs. Without gold, call
it agreement, not accuracy. Stop for review before scaling; a successful pilot
is not permission to label the full population or modify accounts.

`smoke_test=false` is an explicit user waiver, recorded as skipped, never passed.
In simulation mode do not invent a paired API result. Offer real setup or a waiver
and wait. See the reference for copyable prompts and optional task parameters.

## Use safely

The rest: https://github.com/wuyoscar/jev-skill/blob/HEAD/skills/jev-triage/SKILL.md